6949*5174 CASI image of Heron Island 
17 bands, UTM zone 56, ground resolution 1 m, collected July 1, 2 and 3rd 2002
Heron Reef data provided courtesy of the Center for Spatial Environmental Research at the University of Queensland.

Dataset provided by HyPhoon.com.

Please refer to hedley2009.pdf
"Efficient radiative transfer model inversion for remote sensing applications"
John Hedley, Chris Roelfsema, Stuart R. Phinn, 2009


Please also refer to
"Mapping Coral Reef Benthos, Substrates, and Bathymetry,
Using Compact Airborne Spectrographic Imager (CASI) Data", Leiper et al, 2014



work done December 16-19 2013,
from download of data 
to final modeling and upload of results,
including most of this report.


After all those years


 
1 - NO NEED for field data, nor for atmospheric correction
2 - this is demonstrated in this website, using a variety of hyper/multi spectral data
 
Requirements are
1 - homogeneous water body and atmosphere
2 - some coverage of optically deep water
3 - some coverage of dry land
 
Problems are
1 - the precision on estimated depth is found wanting, because the noise-equivalent change in radiance  of accessible data is too high for shallow water column correction work 
2 - radiance data should be preprocessed by the provider at level 1 in order to improve S/N ratio
3 - exponential decay: the deeper/darker the bottom, the poorer the performances
 
So
I keep digging
until suitable data
become available
 

EOMAP
have covered the whole GBR using Landsat archive data (2013)
4SM
offers a real and complete alternative
  • This CASI image of Heron Island is 36 km2. This is 36 million pixels, 17 spectral bands used.
  • At the proposed price of 20 US$ per km2, EOMAP would charge 1294 US$, using free Landsat archive 4 bands, with 30 m ground resolution: this is 40,000 pixels, 4 spectral bands.
  • EOMAP do not disclose their proposed price for finer ground resolution, nor for more spectral bands used.
  • What 4SM offers is that in-house practioners take on this task
    • this would be much more cost-effective,
    • and keep the project "under control" locally,
    • whatever the ground resolution and the number of spectral bands.
  • This would produce the same results: 
    • digital, shallow water bathymetry maps in meters.
    • digital, shallow water seafloor reflectance maps, possibly converted in units of reflectance.
    • an estimate of the water quality: effective spectral diffuse attenuation coefficient.
    • no quality indicator yet: I'm working on it.
  • With the same limitations to try and overcome:
    • Sun glint on the water surface
    • Turbid, or murky waters, preventing detection of seafloor
    • Clouds or haze in the atmosphere
    • Significant waves on the water surface
    • System noise
    • Sensor calibration and radiometric sensitivity
    • Sensor view angle
 home
Workstation
The data
Comment on the data
Image preparation
Preparation of masks
Extraction of calibration data
Optical calibration
Profiles
Modeling
Seatruth
Discrepancies
Downloads
My whishes

next to using a CASI Panchromatic band (march 2014)
4SM Demonstration using this study case (april 2014)




Workstation
  • My computer is a Toshiba Satellite laptop, 4 GB RAM, Intel 3 core processor, running Ubuntu 12.04 LTS
  • I use the PCIDSK image format with FILE interleaving database.pix
  • I use opensource  OpenEV  for display
  • I use a few tools from GMT from University of Hawaii for creation of PostScript plots




The data

 

TCC 644, 547 and 479 nm

FCC 705, 644, and 547 nm
 
 
  • Land areas have been wipped out to zero.
  • A very nice mosaic, nicely balanced.
  • Radiances are reduced to BOA through atmospheric correction.

Bathymetry overlay
derived by University of Queensland


Method used:
see details of the method in Hedley et al's paper (2009) :
quite a complex semi-analytical process indeed






Comment on the data

17 bands used out of 19
WL0439.0/451.2/478.9/498.4/516.1/524.9/547.1/550.3/564.7/585.2/610.0/624.5/643.5/661.0/675.0/690.0/705.0
@ProfileAB/LVW/profile_black/chAB_1_3_16_17

Profile Black
Radiance profile over a deep water area
 
Hedley et al (2009) write:
"The CASI image had been previously corrected to units of above-surface diffuse reflectance, R(0+) Joyce (2004) and so was converted to above-surface remote sensing reflectance using the relation R rs = 0:54R 0þ = kQ where Q = 4 (Mobley, 1994) and k is a spectrally independent factor to convert above-surface to below-surface diffuse reflectance, k ≈ R 0þ = Rð0 − Þ ≈ 0:7, estimated from multiple runs of PlanarRad, an open-source plane-parallel radiative transfer model for directional radiance similar to the commercial software Hydrolight (Hedley, 2008; Mobley, 1994)."

 
 
 
Noise
  • This data exhibits a strong level of noise,
    • which is hardly correlated among bands, if at all.
  • Some glint signal remains present,
    • which we can't remove
    • because of ill correlation among bands.
  • This is not a problem though,
    • as there appears to be no glint inside the lagoon.

Zero
  • I seem to observe that a small amount of atmospheric path has been left.
  • Still some values are at zero: this annoys me: I'm not used to atmospherically corrected data.
Bands 18 and 19
  • They appear to be corrupt: not used here.

Bands 15, 16 and 17
  • They cause distinct under-estimated retrieved depth. Maybe caused by fluorescence in water in the far-red range?
  • They are disabled here.

As this data has been corrected for atmospheric effects, the deep water refelctance in effect provides a measure of the water volume reflectance:
La ~= 0
Lw ~= Lsw



Image preparation
import of data and seatruth  
U. of Queensland's CASI data
  • Export HeronReef_2002_CASI_hyperspectral.dat
    • from ENVI format to PCIDSK format, using OpenEV
  • Create working database using 4SM: 90 channels: see channel descriptor
  • Import bands 1 to 17 into working database, using 4SM, as 16 bits raw data
U. of Queensland's retrieved bathymetry DTM
  • Export HeronReef_2002_BATHY_derived_v00.dat from ENVIformat to PCIDSK format, using OpenEV
    • only an overlay subset is exported
  • Import this overlay into channel 72 of working database
    • Real 32 bits depths in meters are converted into S16 depths in centimeters
Image data may now be inspected and investigated




Masks
@Recode/2_InChannel_18_into_1_OutChannel_18/mask_1
Masks are prepared in channel 18
  • NoData pixels are mapped
  • areas affected by breaking waves are mapped
  • land is masked at 248, including the "void" island
  • the beach is mapped at 21, in order to sample coral sand/rubble to represent  "bright bareland"
  • a sample of deep water area is used for estimation of deep water radiance
 
Now ready
for extraction of calibration data



Extraction of calibration data
-extract/v/RawBDH/FullBDH/NIRband17/NIRmax500/mapBPL/mBPL2
  • This is an automatic process in 4SM: scans the whole image and selects pixels through a set of rules
  • It writes out heron_20020702_m0.cal text file for calibration data (9 MB).
    • this file describes the Brightest Pixels Line for a number of pairs of bands
    • this file describes the Soil Line for a number of pairs of bands
    • and also some other things.
  • It writes out heron_20020702_m0.bdh text file for bi-dimensional histograms
    • this file is even much bigger (167 MB)!
    • this extraction process only takes less than half an hour.



Optical Calibration
-Calibrate/V/BdSMpzg/plot_BPL/BC_17_13_7_3/BDh_12_13_14
Then the practioner engages in the all-important process of optical calibration.
With so many pairs of bands to be considered, it is a fairly cumbersome process.
Only takes an hour or two for a trained person, though.

Calibration diagram for bands 3, 7, 13 and 17
i.e. 479, 547, 643, and 705 nm

Lw=Lsw-La
  • the water volume reflectance Lw has to be estimated from inspection of the calibration diagrams.
  • basic rules of physics of diffusion must be satisfied:
    • Lw is negligeable in the RED to NIR range
    • Lw is significant in the GREEN range
    • Lw can reach fairly high values in clear waters in the Blue range
  • this is very important in respect of dark shallow substrates.
  • it is a question of experience on the part of the practioner.
  • the most important thing is to ensure that the BPL model is parallel to the plot of BPL pixels
    • this is the selection of the brightest pixel in band i for each occurence of radiance in band j
  • BPL pixels are real image pixels
  • BPL model is derived from
  • K3/K7=0.61  provides spectral Kd (see below) without the need/use of any field data, and thereforeallows to compute depths in meters.
    • this is OIB+0.5 marine water type of Jerlov
  • Soil Line and Brightest Pixels Line converge at a very special point which represents the brightest shallow substrate at null depth: LsM
==========Soil Line==========
  • this represents bareland in the image, from bright to dark. Here we only have a small beach made of coral rubbles/sands.
  • it stretches from the LsM point to the La point.
  • it provides a radiance reference for spectrally neutral substrates at null depth, from bright to dark.
==========Lsw==========
  • the deep water radiance Lsw has to be estimated from an optically deep area of the image.
 
The optical calibration of 2Kd from Jerlov's data,
using  observed ratio K3/K7=0.610, WL3=478.9 nm and WL7=547.1 nm
Print_Water_Type in Australia/HeronIsland/heron_20020702.pix is JERLOV type O1B+0.46 for ratio K3/K7=0.610
Ch       1        2        3        4        5        6        7        8        9      10       11      12      13     14      15      16      17 
WL   439.0 459.5 478.9 498.4 516.1 531.1 547.1 564.0 570.9 594.0 610.0 624.5 638.7 665.7 678.9 693.9 705.0  in nm
2K    0.109 0.098 0.095 0.108 0.121 0.136 0.156 0.189 0.204 0.434 0.561 0.645 0.716 0.851 0.943 1.117 2.250  in 1/m
ZM   76.19 85.14 87.59 76.69 68.55 61.37 53.56 44.04 40.95 19.27 14.90 12.96 11.66    9.83  8.89    7.49   3.72  in meters
 
Print_Ratio_K_K at ZM[6]=61.4 m
RATIOS Ki/Kj   K1 /  K2 /  K3 /  K4 /  K5 /  K6 /  K7 /  K8 /  K9 /  K10/  K11/  K12/  K13/  K14/  K15/  K16/  
ZM=76.19  K1  1.000     -     -     -     -     -     -     -     -     -     -     -     -     -     -     - 
ZM=85.14  K2  1.116 1.000     -     -     -     -     -     -     -     -     -     -     -     -     -     - 
ZM=87.59  K3  1.149 1.029 1.000     -     -     -     -     -     -     -     -     -     -     -     -     - 
ZM=76.69  K4  1.008 0.903 0.877 1.000     -     -     -     -     -     -     -     -     -     -     -     - 
ZM=68.55  K5  0.900 0.806 0.783 0.893 1.000     -     -     -     -     -     -     -     -     -     -     - 
ZM=61.37  K6  0.805 0.721 0.700 0.799 0.895 1.000     -     -     -     -     -     -     -     -     -     - 
ZM=53.56  K7  0.701 0.628 0.610 0.696 0.779 0.871 1.000     -     -     -     -     -     -     -     -     - 
ZM=44.04  K8  0.577 0.517 0.502 0.573 0.642 0.717 0.823 1.000     -     -     -     -     -     -     -     - 
ZM=40.95  K9  0.536 0.480 0.466 0.532 0.596 0.666 0.764 0.928 1.000     -     -     -     -     -     -     - 
ZM=19.27  K10 0.252 0.225 0.219 0.250 0.280 0.313 0.359 0.436 0.470 1.000     -     -     -     -     -     - 
ZM=14.90  K11 0.195 0.174 0.169 0.193 0.216 0.242 0.278 0.337 0.363 0.773 1.000     -     -     -     -     - 
ZM=12.96  K12 0.169 0.152 0.147 0.168 0.188 0.210 0.242 0.293 0.316 0.673 0.870 1.000     -     -     -     - 
ZM=11.66  K13 0.152 0.137 0.133 0.151 0.170 0.189 0.218 0.264 0.285 0.606 0.784 0.901 1.000     -     -     - 
ZM= 9.83  K14 0.128 0.115 0.112 0.127 0.143 0.159 0.183 0.222 0.240 0.510 0.660 0.758 0.842 1.000     -     - 
ZM= 8.89  K15 0.116 0.104 0.101 0.115 0.129 0.144 0.165 0.201 0.216 0.461 0.596 0.684 0.760 0.903 1.000     - 
ZM= 7.49  K16 0.098 0.088 0.085 0.097 0.109 0.122 0.140 0.170 0.183 0.389 0.503 0.578 0.642 0.762 0.844 1.000 
ZM= 3.72  K17 0.049 0.043 0.042 0.048 0.054 0.060 0.069 0.084 0.091 0.193 0.250 0.287 0.318 0.378 0.419 0.496